Distinguishing nontuberculous mycobacterial lung disease and Mycobacterium tuberculosis lung disease on X-ray images using deep transfer learning

Table 5 Three CNN models and the radiologist results for the test set

Class	Model	Precision	Recall	F1 score	AUROC	Accuracy	Average IOBB
NTM-LD (NP = 26, NI = 26)	DenseNet 201	0.61	0.77	0.68	0.82	0.77	0.24
	ResNet 50	0.59	0.85	0.70	0.85	0.85	0.27
	EfficientNet B4	0.71	0.85	0.77	0.88	0.85	0.36
	Radiologist	0.58	0.73	0.64	0.80	0.73	N/A
MTB-LD (NP = 108, NI = 108)	DenseNet 201	0.94	0.88	0.91	0.82	0.88	0.35
	ResNet 50	0.96	0.86	0.91	0.85	0.86	0.39
	EfficientNet B4	0.96	0.92	0.94	0.88	0.92	0.50
	Radiologist	0.93	0.87	0.90	0.80	0.87	N/A

Bold text means the highest performance, NP: Number of Patients, NI: Number of Images.
In the case of a radiologist in IOBB, it is used as ground truth and is marked as (N/A).